With the increasingly complex tracking targets and environment in radar application, the widely used Kalman filter and extended versions can track targets with less few calculations. However, it requires the system noises including process noise and measurement noise in a zero mean and Gaussian case, otherwise, it may lead to poor performance. In order to solve the problem of tracking a maneuvering target, when process and measurement noises are unknown-but-bounded, an extended set-membership filter and the tracking quality method based on information geometry are proposed. Simulation analysis shows that the set-membership filter provides distinct advantages in tracking performance including state bounding and tracking accuracy, and the maneuver behaviors of target can be identified through counting the changes of distances on the manifold.